基于深度学习的急性白血病流式细胞术检测报告文本资料自动分类研究  

Automatic Classification of Text Data for Flow Cytometry Detection of Acute Leukemia Based on Deep Learning

作　　者：张亚洲 李智伟[2] 农卫霞[3] 雷伟 摆文丽 李寅臻 李瑞 王奎 ZHANG Yazhou;LI Zhiwei;NONG Weixia;LEI Wei;BAI Wenli;LI Yinzhen;LI Rui;WANG Kui(Department of Preventive Medicine,Shihezi University School of Medicine,Shihezi 832000,Xinjiang,China;Clinical Testing Center,People's Hospital of Xinjiang Uygur Autonomous Region,Urumqi 830001,Xinjiang,China;Department of Rheumatology and Hematology,Shihezi University School of Medicine,Shihezi 832000,Xinjiang,China)

机构地区：[1]石河子大学医学院预防医学系,新疆石河子832000 [2]新疆维吾尔自治区人民医院临床检测中心,新疆乌鲁木齐830001 [3]石河子大学医学院风湿血液科,新疆石河子832000

出　　处：《医学信息》2025年第8期15-20,共6页Journal of Medical Information

基　　金：国家自然科学基金项目(编号:81860374)。

摘　　要：目的 探索深度学习模型在流式细胞检测报告结果部分文本资料上的分类效果。方法 使用CNN、LSTM等六种深度学习模型对流式检测报告的结果部分的文字资料进行分析,并对急性白血病患者进行分类预测,最后通过综合指标F1值对模型进行评价。结果 CNN-BiLSTM混合模型的精确率、召回率、F1值最优,分别为0.7422、0.7365、0.7361;模型在正常人、急性髓系白血病、急性B淋巴细胞白血病、有核红细胞异常、中性粒细胞异常、浆细胞异常、单核细胞异常这7类的F1值均达到了70%。结论 混合模型对流式细胞术检测报告结果部分的文本资料的分类效果较好,可与前期研究联合共同构建了一个更为完整的流式细胞术自动化分析体系,进一步提高流式细胞术分析的效率和准确性。Objective To explore the classification effect of deep learning model on text data of flow cytometry report results.Methods Six deep learning models such as CNN and LSTM were used to analyze the text data of the results of the flow cytometry report,classify and predict the patients with acute leukemia,and finally evaluate the model by the comprehensive index F1 score.Results The precision,recall and F1 score of the CNNBiLSTM mixed model were the best,which were 0.7422,0.7365 and 0.7361,respectively,and the F1 score of the model reached 70% in seven categories:normal humans,acute myeloid leukemia,acute B lymphoblastic leukemia,nucleated red blood cell abnormalities,neutrophil abnormalities,plasma cell abnormalities and monocytic abnormalities.Conclusion The mixed model has a good effect on the classification of text data in the results of flow cytometry test report,and can be combined with previous studies to build a more complete automated flow cytometry analysis system to further improve the efficiency and accuracy of flow cytometry analysis.

关 键 词：流式细胞术 文本分类 CNN 自动化分析 深度学习 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]